[libvirt] [PATCH v3] interfaces: Convert virInterfaceObjList to virObjectRWLockable

[John Ferlan]

[...]

>>>
>>> With the patch split in 2 introducing 2 distinct changes + the NULL check:
>>> Reviewed-by: Erik Skultety <eskultet at redhat.com>
>>
>> Hopefully you reconsider the desire for 2 patches...
> 
> Well, since I was apparently fine with the change when reviewing the same
> changes to nodedev, I guess I should be fine with it now too, I don't know what
> made me change my opinion as time has passed, nevermind, it's not a deal breaker
> for me.
> 

OK - thanks - I did note while looking deeper that I didn't remove
virInterfaceObjListFree() from src/conf/virinterfaceobj.h, so I removed
it...

>>
>>>
>>>
>>> PS: I'm also sad, that we have two backends here just like we have in nodedev
>>> with one of them being essentially useless (just like in nodedev) we have this
>>> 'conf' generic code (just like in nodedev), yet in this case it's only used in
>>> the test driver. I'd very much appreciate if those backends could be adjusted
>>> in a way where we could make use of these functions. I can also imagine a
>>> cooperation of the udev backend with the nodedev driver where we have an active
>>> connection to the monitor, thus reacting to all events realtime, instead of
>>> defining a bunch of filters and then letting udev re-enumerate the list of
>>> interfaces each and every time (and by saying that I'm also aware that udev is
>>> actually the useless backend here).
>>>
>>> Erik
>>>
>>
>> Yeah - the driver code here is quite different/strange and could
>> possibly use a bit more convergence. I feel too battered and bruised
>> over this convergence right now though ;-)....  Besides the differences
> 
> I hope it didn't sound like a request, it was meant to be more like a wish that
> we should do something about it (sometime).
> 
> Erik
> 

I didn't take it that way... Just making sure no one would be waiting
for patches from me any time soon that did that type of convergence ;-)

Tks -

John